Article Text


S57 Diabetes and latent tuberculosis infection: nested case-control study within the PREDICT cohort
  1. C Jackson1,
  2. J Southern2,
  3. HS Whitworth3,
  4. M Scott3,
  5. C-Y Tsou4,
  6. S Sridhar3,
  7. V Nikolayevskyy5,
  8. M Lipman1,
  9. A Sitch6,
  10. J Deeks6,
  11. C Griffiths5,
  12. F Drobniewski4,
  13. A Lalvani3,
  14. I Abubakar1,7 On behalf of
  1. 1University College London, London, UK
  2. 2Health Protection Services, Public Health England, London, UK
  3. 3Imperial College London, London, UK
  4. 4National Mycobacterium Reference Laboratory, Public Health England, London, UK
  5. 5Queen Mary University of London, London, UK6; University of Birmingham, Birmingham, UK
  6. 7The PREDICT Study Group


Background Diabetes is associated with an increased risk of tuberculosis disease, but it is unclear whether a similar association exists between diabetes and latent tuberculosis infection (LTBI).

Methods The ongoing UK PREDICT (Prognostic Evaluation of Diagnostic IGRAs Consortium) cohort study aims to recruit 10,000 participants to assess the predictive values of interferon gamma release assays (IGRAs) for the development of active TB in recent entrants to the UK and contacts of active TB cases. We used a nested case-control design within the first 5000 recruits in this cohort, to investigate the association between diabetes and LTBI. Participants in PREDICT provide demographic, medical and social information, including any history of diabetes. LTBI is detected using the two commercially available IGRAs, Quantiferon Gold In-Tube and TSpot.TB. Cases were individuals who tested positive on either or both IGRAs; controls were negative on both assays (or negative on one and indeterminate on the other). Logistic regression was used to estimate odds ratios and adjust for potential confounders. Assuming a 5% diabetes prevalence, 1084 cases and 3252 controls would allow the detection of a 1.5-fold increase of LTBI with 80% power and 5% error.

Results Overall, 1388/4730 (29%) had a positive IGRA. 286/4730 (6%) reported a history of diabetes. Amongst diabetic participants, 168 used insulin and/or oral hypoglycaemic medications and 25 reported control through diet alone (1 participant was being monitored only and for 92 the level of control was unknown). Univariate analysis found an association between diabetes and LTBI (OR = 1.45 [95% CI 1.13–1.86], p = 0.003). After adjustment for age, this association was no longer apparent (OR = 1.15 [95% CI 0.88–1.50], p = 0.30). Adjustment for other variables in addition to age (sex, ethnicity, birthplace outside the UK, previous contact with a TB case, or previous TB diagnosis) did not substantially change the estimated age-adjusted OR relating diabetes to LTBI. Similar results were obtained when the analysis was restricted to contacts.

Conclusions In the current PREDICT cohort, diabetes does not appear to be associated with LTBI after adjustment for age. The relationship between diabetes and TB disease observed elsewhere may reflect an increased risk of disease rather than infection.

Statistics from

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.